Download Technics SA-DX950 Receiver

Hearing Research 202 (2005) 235–247
www.elsevier.com/locate/heares
Attribute capture in the precedence effect
for long-duration noise sounds
Liang Li
a,b,*
, James G. Qi b, Yu He c, Claude Alain c, Bruce A. Schneider
b
a
b
Department of Psychology, Speech and Hearing Research Center, National Key Laboratory on Machine Perception,
Peking University, Beijing 100871, China
Centre for Research on Biological Communication Systems, Department of Psychology, University of Toronto at Mississauga,
3359 N Mississauga Road, Mississauga, Ont., Canada L5L 1C6
c
The Rotman Research Institute, Baycrest Centre for Geriatric Care, Toronto, Ont., Canada M6A 2E1
Received 7 October 2004; accepted 13 October 2004
Available online 8 December 2004
Abstract
Listeners perceptually fuse the direct wave from a sound source with its reflections off nearby surfaces into a single sound image,
located at or near the sound source (the precedence effect). This study investigated how a brief gap presented in the middle of either a
direct wave or simulated reflection is incorporated into the fused image. For short (<9.5 ms) delays between the direct (leading) and
reflected (lagging) waves, no sound was perceived from the direction of the lagging wave. For delays between 10 and 15 ms, both
sounds were perceived, but the gap was heard only on the leading side. When the gap was only in the correlated lagging sound at
short delays, it also was perceived as occurring on the leading side. Moreover, gap detection thresholds were the same for gaps in the
leading and lagging sounds, suggesting that the perception of the gap was not suppressed, but rather incorporated into the leading
sound. Finally, scalp event-related potentials were not associated with the precedence effect until the gap occurred. This suggests that
cortical mechanisms are engaged to maintain fusion when attributes in direct or reflected waves change.
2004 Elsevier B.V. All rights reserved.
Keywords: Precedence effect; Fusion; Reverberant environment; Correlation; Gap; Event-related potential
1. Introduction
In a reverberant environment, each sound source produces both a direct wavefront and numerous filtered and
time-delayed reflections from the walls, ceilings and
other surfaces. When the delay between the direct wave
and a reflected wave is sufficiently long and the reflected
Abbreviations: B&K, Bru¨el & kjær; ERP, event-related potential;
HATS, head and torso simulator; IAC, Industrial Acoustic Company;
RO, right loudspeaker was turned on only; L/U, left leading/uncorrelated; L/C, left leading/correlated; R/C, right leading/correlated;
TDT, Tucker–Davis technologies
*
Corresponding author. Tel.: +905 569 4628; fax: +905 569 4326/1
416 978 4811.
E-mail address: [email protected] (L. Li).
0378-5955/$ - see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.heares.2004.10.007
wave is sufficiently intense, the reflected wave is perceived as a distinct auditory event (an echo), whose perceived location is usually different from that of the
source. However, when the delays between the direct
wavefront and its reflections are short (e.g., 1–10 ms
or more, depending on the stimulus), the auditory system somehow gives ‘‘precedence’’ to the direct wavefront over its reflections so that the listener hears only
a single fused sound whose point of origin is perceived
to be at or near the location of the sound source. This
phenomenon is called the ‘‘precedence effect’’ (Clifton
and Freyman, 1989; Freyman et al., 1991; Shinn-Cunningham et al., 1993; Wallach et al., 1949; Zurek, 1980;
for reviews see Blauert, 1997; Li and Yue, 2002; Litovsky et al., 1999; Zurek, 1987).
236
L. Li et al. / Hearing Research 202 (2005) 235–247
The precedence effect reduces listenersÕ perception of
multiple images by perceptually grouping correlated
acoustic waveforms from different directions, thereby
avoiding the perception of multiple sound images when
only one source is present. Furthermore, because the
fused image is perceived as originating at or near the
location of the source, localization errors are reduced
in reverberant environments. In experimental environments, the ‘‘direct’’ and ‘‘reflected’’ waves are usually
produced by two spatially separated sound sources,
and the shortest time delay between a direct and a reflected wave that produces a separate echo on certain
percentage of experimental trials (usually between 50%
and 80%) is called the echo threshold (Blauert, 1997,
pp. 224–225).
Since a simulated reflection in an experimental environment is not heard as a separate auditory event when
the lead/lag delay is below the echo threshold, it has
been assumed that some inhibition or attenuation of
information in reflected sounds, such as contralateral
inhibition (Blauert, 1997, pp. 230–233), must take place
in the precedence effect. For instance, a prevalent explanation is that the directional information associated
with the reflected wave is suppressed (Blauert, 1997; Liebenthal and Pratt, 1999; Litovsky and Shinn-Cunningham, 2001; Rakerd et al., 2000; Yin, 1994; Zurek,
1980). This suppression hypothesis has dominated the
search for neural correlates of the precedence effect. In
most of the related physiological studies using either
anesthetized or unanesthetized animals, suppressed neural responses to the lagging sound in the presence of the
leading sound were treated as the neural correlates of
the precedence effect (Fitzpatrick et al., 1995, 1999; Liebenthal and Pratt, 1999; Litovsky, 1998; Litovsky and
Delgutte, 2002; Litovsky and Yin, 1998a,b; Litovsky
et al., 1997; Yin, 1994).
However, suppression of the directional information
in the reflection does not mean that the reflected wave
is not heard because listeners are aware of the presence
of reflections and even changes in them. For example,
Freyman et al. (1998) have shown that listeners are as
sensitive to intensity decreases in the lagging sound as
to intensity increases in the leading sound, indicating
that intensity information in the reflection is not suppressed. Also, hearing a reflection while presumably
suppressing its directional information raises some puzzles as to how the perceptual system incorporates reflected waves into the percept of a single auditory
event. For example, it is not clear how the intensities
of a source and its reflections blend to determine the
loudness of the ‘‘fused’’ sound image. Finally, Hartung
and Trahiotis (2001) have developed a model for
describing how monaural peripheral processing without
an inhibitory mechanism may contribute to data obtained in binaural ‘‘precedence’’ experiments that use
binaural pairs of transients as stimuli. Hence, it is evi-
dent that there is more to the precedence effect than simple inhibition.
Most studies on the precedence effect have used idealized brief acoustic stimuli, such as clicks or transient
noise bursts, to avoid or reduce temporal overlap between the leading and lagging sounds (for a review see
Litovsky et al., 1999). However, acoustic stimuli under
normal circumstances are usually complex and last for
more than a few hundred milliseconds. Therefore, it is
important to study how the precedence effect works
for long-duration stimuli, and determine how attributes
that belong to reflections, and indeed may be unique to
them, are incorporated into the fused image of the
source.
In the present study, a transient gap, as a probe attribute, was inserted into an otherwise continuous steadystate broadband noise. Because this gap could be in the
source (the leading sound) only, the reflection (the lagging sound) only, or both source and reflection, it
should be easier to determine how this attribute of the
direct wave and/or the reflection is detected and incorporated in the overall percept of the sound.
Introducing a single gap into either the leading or the
lagging sound (but not both) is also interesting from the
point of view of top-down control over the precedence
effect. For example, a gap only in the lagging but not
in leading stimulus is inconsistent with the lagging stimulus being an echo (a gap in a natural reflection should
have its origin in the sound source), and could lead to a
breakdown in the precedence effect. Moreover, if the gap
is in the lagging stimulus only, and the leading and lagging stimuli remained fused into a single percept, will the
listener perceive a break in the fused stimulus, or will the
gap in the lagging stimulus be suppressed so that the listener hears a continuous fused stimulus? To investigate
issues such as these, listeners were asked to describe their
experience to the gap, which was introduced into the
middle of either the leading or lagging sound.
As mentioned earlier, most neurophysiological studies on the precedence effect have mainly focused their efforts on determining the brainstem mechanisms
involved in lag suppression in experimental animals
(Fitzpatrick et al., 1995, 1999; Litovsky, 1998; Litovsky
and Delgutte, 2002; Litovsky and Yin, 1998a,b; Litovsky et al., 1997; Yin, 1994). However, there is more to
precedence than simple suppression of the location
information of the lagging stimulus. For example, several studies have shown that listenersÕ knowledge and
expectations about the room acoustics can strongly affect the precedence effect (Clifton, 1987; Clifton and
Freyman, 1989; Clifton et al., 1994; Freyman et al.,
1991). Repeated presentations of the leading and lagging
clicks, which are not perceived to be fused at the beginning, can eventually cause fusion to occur, suggesting
that following continued exposure to a reverberant environment, listeners can build up a new representation of
L. Li et al. / Hearing Research 202 (2005) 235–247
the room acoustics consistent with the leading and lagging stimulus being produced by a single source. Moreover, once fusion is established, it is most readily broken
when a change in the spatial relationship between the
leading and lagging sounds is inconsistent with the
knowledge of the room acoustics that has been acquired
previously. Thus there is a strong higher-order cognitive
component involved in the precedence effect. For this
reason, humanÕs cortical correlates of the precedence effect were investigated using the method of scalp eventrelated potential (ERP) recording. Since ERPs to a brief
acoustic event can last a few hundred ms, in the present
study, the sound duration was set to about 3 s so that
ERP responses specific to the probe gap could be more
easily separated from those to sound onset and offset.
2. Experiment 1
In the first experiment, echo thresholds for long-duration noises as a function of the delay between the direct
wave and its simulated reflection were measured. This
threshold is defined as the longest delay between the direct and reflected wave at which no sound is perceived
from the direction of the lagging stimulus. A gap capture threshold was also determined, where the gap capture threshold is defined as the longest delay between the
direct and reflected wave at which the listener could no
longer detect a gap in stimulation from the direction of
the lagging sound.
2.1. Materials and methods
2.1.1. Participants
Fifteen young (19–25 years old, six females and nine
males) university students with normal and balanced
(less than 15 dB difference between the two ears) puretone hearing, confirmed by audiometry, participated in
this experiment. The audiometric thresholds were determined at frequencies of 250, 500, 1000, 2000, 3000, 4000,
6000, and 8000 Hz and all listeners in this and in subsequent experiments were normal for frequencies less than
6 kHz. All the listeners in this and next experiments gave
their written informed consent to participate in the
experiments and were paid a modest stipend for their
participation.
2.1.2. Apparatus and materials
During test sessions, listeners were seated in a chair at
the center of an Industrial Acoustic Company (IAC)
sound-attenuated chamber, whose internal dimensions
were 193 cm in length, 183 cm in width, and 198.5 cm
in height. Gaussian broadband noises (0–10 kHz),
whose duration was 3050 ms (including 30 ms rise-fall
times), were synthesized using a 16-bit Tucker–Davis
Technologies (TDT) System II hardware DD1 and cus-
237
tom software at the sampling rate of 20 kHz. The noise
signals were converted to analog forms using TDT DD1
digital-to-analog converters under the control of a Dell
computer with a Pentium processor. The analog outputs
were low-passed at 10 kHz with the TDT FT5 filter,
attenuated by two programmable attenuators (TDT
PA4, for the left and right channels), amplified via a
Technics power amplifier (SA-DX950), and then delivered from two balanced loudspeakers (Electro-Medical
Instrument, 40 W), which were in the frontal azimuthal
plane at the left and the right 45 positions symmetrical
with respect to the median plane (Fig. 1). The distance
between each of the two loudspeakers to the center of
the listenersÕ head was 1.03 m. The loudspeaker height
was approximately ear level for a seated listener with
the average body height. Fresh noise sounds were generated for each trail. The gap, a rectangular silent break in
the otherwise continuous noise, occurred 1500 ms after
sound onset.
When the noises delivered to the two loudspeakers
were identical (except for a delay between them),
they were referred to as ‘‘correlated’’. When the two
noises were independent, they were referred to as ‘‘uncorrelated’’. All the single-source levels were fixed at 60 dB
SPL. Calibration of sound level was carried out with a
Bru¨el & Kjær (B&K) sound meter (Type 2209) whose
microphone was placed at the location of the listenersÕ
head center when the listener was absent. ‘‘A’’ weighting
and a ‘‘slow norm’’ meter response were used.
2.1.3. Procedure
There were three stimulus conditions in this experiment. In Condition Correlated/No-gap, the two noises
from the two loudspeakers were correlated and no gaps
0o
-45o
+45o
Fig. 1. Diagram showing the two-loudspeaker configuration used in
the present study. The two loudspeakers were spatially separated in the
frontal azimuthal plane at the left and the right 45 positions
symmetrical with respect to the median plane.
238
L. Li et al. / Hearing Research 202 (2005) 235–247
were introduced. The left loudspeaker led the right loudspeaker, and the time lag between them (called the lead/
lag time) was reduced following responses indicating a
perceived noise sound from the right loudspeaker, and
increased following responses indicating that no noise
sound was perceived from the right loudspeaker using
a 3-down-1-up procedure (Levitt, 1971). All sessions
were started with a 50 ms lead/lag time. Therefore, the
longest lead/lag time, at which no sound image from
the right loudspeaker was perceived (the ‘‘echo inaudible’’ criterion), was obtained. That an echo (not a reflection) is perceived or not is subjective, and listenersÕ
responses cannot be categorized as either ‘‘correct’’ or
‘‘incorrect’’. Thus in this and the other two conditions
of this experiment, no feedback was given to listeners.
In Condition Correlated/50-ms-gap, a 50-ms gap was
introduced into the middle of each of the two correlated
noises from the two loudspeakers. The delay between
the onsets of the two gaps was equal to the delay between the leading and lagging sounds. The left loudspeaker was also the leading loudspeaker, and
listeners, when presented with a stimulus, indicated by
pressing one of two buttons whether they heard a gap
in the sound coming from the right (lagging) source.
Logically, of course, they could only hear a gap in the
right-side noise if they heard a noise on the right. Hence,
the question here is whether, when they heard a noise on
the right (lead/lag delays > echo threshold), they also
perceived a gap in the right-side noise, or whether the
gap was only heard in the left-side (leading) noise. If
they did not hear a gap in the noise coming from the
right loudspeaker they were to press the other button.
In other words, the lag time between the sounds from
the two loudspeakers was reduced following responses
indicating a perceived gap in the noise perceived on
the right, and increased following responses indicating
that they did not hear a gap on the right. The same 3down-1-up procedure was employed (Levitt, 1971).
In Condition Uncorrelated/50-ms-gap, a 50-ms gap
was introduced into the middle of each of the two uncorrelated noise sounds from the two loudspeakers, and the
procedure was the same as that of Condition Correlated/
50-ms-gap. There were four repetitions in each of the
three conditions.
individualsÕ echo thresholds, only one noise sound image
was heard as coming from the locus of the leading loudspeaker and no sound image as coming from the right
loudspeaker was perceived. As shown in Fig. 2, the average echo threshold was approximately 9.5 ms.
When a gap was introduced into both the leading and
lagging sounds in Condition Correlated/50-ms-gap, the
average gap capture threshold was 15.6 ms (Fig. 2).
The gap capture threshold in Condition Correlated/50ms-gap was significantly longer than the echo threshold
in the same condition (F1,14 = 5.769, MSE = 47.617,
p = 0.031). At delays substantially longer than the gap
capture threshold, listeners perceived a gap in the sound
image associated with the right loudspeaker. At delays
between the echo threshold, and the gap capture threshold, listeners perceived sounds from both the left (leading) and right (lagging) loudspeakers, but did not hear a
gap in the lagging sound. Rather the gap was heard only
in the leading sound. Finally, at delays shorter than the
echo threshold listeners only heard a sound on the left
with a gap in it. Hence, for intermediate delays (between
10 and 15 ms) in Condition Correlated/50-ms-gap, listeners heard two spatially separated continuous sound
images (a direct wave and its echo), with a gap in the
leading image, but not in the lagging image, even though
both leading and lagging sounds contained a 50 ms gap.
In Condition Uncorrelated/50-ms-gap, listeners always perceived two spatially distinct sounds (one on
the left and the other on the right), regardless of the
2.2. Results
In Condition Correlated/No-gap, when the lead/lag
times were substantially longer than the individualsÕ
echo thresholds, all 15 listeners perceived a distinct
sound image originating from the right loudspeaker. Because the noise sound image originating from the left
loudspeaker was always perceived, two spatially separate noise sounds were actually heard at the longer
lead/lag times, one on the left and one on the right.
When the lead/lag delays were substantially below the
Fig. 2. Comparison of average attribute capture thresholds between
the two conditions: (1) Condition Correlated/No-gap, the two noises
from the two loudspeakers were correlated and no gaps were
introduced. (2) Condition Correlated/50-ms-gap, a 50-ms gap was
introduced into the middle of each of the two correlated noise sounds
from the two loudspeakers. The error bars indicate the standard errors
of the mean.
L. Li et al. / Hearing Research 202 (2005) 235–247
lead/lag time. Thirteen of the 15 listeners always heard
gaps in both sounds at all delays, however, two of the
listeners occasionally reported that they did not hear a
gap in the lagging sound.
3. Experiment 2
In Experiment 1, when there were gaps in both leading and lagging correlated noises, and the lead/lag time
was slightly longer than the echo threshold (10–15 ms),
so that both the leading and lagging sounds were heard,
listeners heard a gap in the leading but not in the lagging
sound. A possible explanation of this phenomenon is
that some attributes of the lagging sound (e.g., the presence of a gap) were being suppressed, even though the
lagging sound was heard. If that were the case then it
would be expected that attributes of the lagging sound
would be even more suppressed when the lead/lag time
was short enough that only a single fused sound was
heard. To check whether attributes of the gap were suppressed when the lagging sound was clearly captured, in
the second experiment gap detection thresholds (the
shortest duration at which a gap was perceived), both
when sounds were fused (echo capture) and when they
were not, were determined.
To see whether a listenerÕs sensitivity to a gap depended on whether or not fusion occurred, in Experiment 2 gap detection thresholds when fusion clearly
happen (correlated noises, 2 ms delay) were compared
to a condition when it did not (uncorrelated noises, 2
ms delay). If the gap appeared only on the lagging side
and was suppressed when fusion occurred, then the
gap detection threshold should be higher than when
there was no fusion.
3.1. Materials and methods
3.1.1. Participants
The fifteen people who participated in Experiment 1
also participated in this experiment.
3.1.2. Apparatus and materials
The apparatus and materials were same as in
Experiment 1.
3.1.3. Procedure
Unlike Experiment 1, where there was a gap in the
noises produced by both the left and right loudspeakers,
in Experiment 2, the gap appeared only in the noise that
was delivered from the right loudspeaker. The minimum
size of the gap in the right-loudspeaker noise that could
be detected using a single-interval staircase procedure
was then determined, for both correlated and independent leading and lagging noises. Specifically, on each
trial a stimulus with a gap in the sound emanating from
239
the right loudspeaker was presented. If the listener responded that she/he heard the gap on three consecutive
trials, the duration of the gap on the next trial was reduced. If, however, the listener indicated on a trial that
they could not hear a gap, the duration of the gap on the
next trial was increased, a 3-down (gap duration reduced)-1-up (gap duration increased) procedure (Levitt,
1971). 1
In Condition RO, the right loudspeaker was turned
on and the left loudspeaker was turned off. In Condition
L/U, a right-side noise sound (with a gap) lagged 2 ms
behind an uncorrelated left-side noise sound without a
gap. 2 In Condition L/C, a right-side noise sound (with
a gap) lagged 2 ms behind a correlated left-side noise
sound without a gap. In Condition R/C, a right-side
noise sound (with a gap)led, by 2 ms, a correlated leftside noise sound without a gap. There were four repetitions in each of the conditions. The maximum gap at the
beginning of a session was 50 ms.
3.2. Results
As indicated in Fig. 3, the gap detection thresholds
among Conditions L/U, L/C and R/C were similar,
and the lowest gap detection threshold was obtained
when only the right loudspeaker was operative (Condition RO). A one-way analysis of variance with repeated
measures revealed that the differences in gap detection
thresholds between these four conditions were significant (F3,42 = 5.146, MSE = 6.030, p = 0.004). Pairwise
analyses indicated that Condition RO was significantly
different from each of the other three conditions
(p < 0.005) but there were no significant differences
among Conditions L/U, L/C and R/C (p > 0.800). Hence
1
We opted to use a single-interval staircase procedure rather than
the more standard two-interval, forced-choice procedure for two
reasons. First, the use of a two-interval technique would have more
than doubled trial length from its current 3.05 s to more than 7 s (once
an inter-stimulus interval was added), and we were concerned about
tiring our volunteers. Second, we wanted to keep the testing situation
as comparable as possible to that used in Experiment 1 (where we also
used a single-interval staircase procedure) since we were using naı¨ve
listeners. Although thresholds determined using single-interval staircase procedures are subject to response biases, such biases are not a
significant problem for comparisons of thresholds as long as these
biases remain constant across comparisons. Because there is no reason
to expect that a change from left leading to right leading, or from
correlated to independent noises, or from the left loudspeaker ‘‘on’’ to
the left loudspeaker ‘‘off’’ would affect the bias to report a gap, gap
detection threshold differences among these conditions should accurately reflect relative (but perhaps not absolute) sensitivity to the
presence of a gap.
2
Because the two uncorrelated sounds did not fuse, it should not
matter whether right or left was leading for detecting the gap in the
middle of the right sound, especially when the gap occurred 1500 ms
after sound onset. Thus for the gap detection test, Condition R/U
should be equivalent to Condition L/U and was not included in the
experimental protocol.
240
L. Li et al. / Hearing Research 202 (2005) 235–247
Fig. 3. Comparison of average gap detection thresholds in the
following four conditions: (1) right sound only (RO), (2) left leading/
uncorrelated (L/U), (3) left leading/correlated (L/C), and (4) right
leading/correlated (R/C). The error bars indicate the standard errors of
the mean.
there was no indication that changing the left noise from
lagging to leading affected the detection of a gap. Indeed, the gap detection threshold remained unchanged
even when the two sounds were uncorrelated. These results are consistent with the notion that the detection of
a gap in a stimulus depends only on the extent of the
drop in acoustic energy present in the stimulus at
the ears, since the degree of interaural correlation and
the direction of the lag apparently had no effect on
threshold.
To determine the nature of the local cues in the left
and right ear that could signal the presence of a gap, a
B&K head and torso simulator (HATS, 4128C) was
placed at the position that would be occupied by the listenersÕ head. The signal at the location of the eardrum in
the simulated head was then recorded for both left and
right ears under two conditions using the B&K Pulse
Platform. In the first condition, correlated noises were
presented over both loudspeakers with the left loudspeaker leading the right loudspeaker by 2 ms. The lines
with filled circles in Fig. 4 depict the long-term spectra of
the left- (left panel) and right- (right panel) ear signals
when both loudspeakers were playing. The lines with
open squares depict the long-term spectra of the left (left
panel) and right (right panel) ear signals when only the
left loudspeaker was on (i.e., the condition that existed
when there was a gap in the right loudspeaker). The differences between the two spectra in the left panel identify the left-ear spectral cues to the presence of a gap.
The comparable differences in the right panel identify
the right-ear spectral cues to the presence of a gap.
Clearly, spectral differences in the right ear are much
more pronounced than they are in the left ear, especially
at the high frequencies (due to the head shadow effect).
Because gap detections thresholds did not vary with the
degree of interaural correlation, and because the spectral
cues are much more pronounced at the right ear, it is
reasonable to conclude that the detection of a gap was
based on the processing of intensity information in the
right ear. Hence, on the basis of spectral cues, it would
Right Ear Canal
0
-5
-10
-15
-20
-25
-30
-35
-40
-45
-50
-55
-60
RELATIVE POWER IN dB
RELATIVE POWER IN dB
Left Ear Canal
0
1
2
3 4 5 6 7 8
FREQUENCY IN kHz
9 10
0
-5
-10
-15
-20
-25
-30
-35
-40
-45
-50
-55
-60
0
1
2
3
4
5
6
7
8
9 10
FREQUENCY IN kHz
Fig. 4. Long-term spectra for the stimuli in this experiment when the listener was replaced by simulated head and torso (Bruel & Kjaer). The lines
connecting the open squares represent the spectra of sounds in the two ear canals of the simulated head when the noise stimulus was being played
over the left loudspeaker only. The line connecting the filled circles represents the spectra of the sounds in the two ear canals when the same sound
was presented over both left and right loudspeakers with the sound on the left leading that on the right by 2 ms. The left panel presents these spectra
for the left ear canal; the right panel for the right ear canal.
L. Li et al. / Hearing Research 202 (2005) 235–247
be expected that the listener might hear the gap as occurring on the right. However, for gap durations just above
threshold, listeners reported (after the session) that they
perceived a gap in the sound source located to the left.
In other words, the leading sound appeared to fully
capture an attribute in the lagging sound. This capture
effect was explored more systematically in the next
experiment.
4. Experiment 3
Experiment 3 investigated how the precedence effect
modified listenersÕ perceptions of a gap that appeared
either in the lagging or leading sounds, but not both.
Specifically, listeners were asked to report their impressions associated with gaps in Conditions L/U, L/C,
and R/C (see Experiment 2 for the definitions of the
three conditions).
241
speaker(s) delivered the perceived noise burst(s) (for
the instructions to listeners, see Footnote 3). Thus Options 1 and 2 were associated with perception of only
one brief auditory event in the middle of the noise
sound, and Options 3, 4, and 5 were associated with perception of 2 brief auditory events. Option 6 indicated
that the participant did not perceive any event in the
middle of the noise.
Noise burst options were incorporated into the response list because there were reasons to expect that listeners would hear a noise burst if there was any
tendency for echo capture to break down during a
gap. For example, if a gap were introduced into the leading stimulus only, there would be no leading stimulus
during the gap to suppress the information as to the
location of the lagging stimulus. Hence, one might expect to hear a brief noise burst from the location of
the lagging stimulus.
4.2. Results
4.1. Materials and methods
4.1.1. Participants
Eleven listeners (four females and seven males) with
normal and balanced pure-tone hearing participated in
this experiment. Four young male listeners also participated in Experiments 1 and 2. The other 7 listeners included 4 young female listeners (19–31 years old), and
3 male listeners (34, 34 and 39 years old, respectively).
The gap detection threshold for each of these 7 listeners,
who did not participated in Experiments 1 and 2, was
also measured under Condition L/U.
4.1.2. Apparatus and materials
The apparatus and materials were the same as in
Experiments 1 and 2.
4.1.3. Procedure
Stimuli were presented in each of the three conditions
(L/U, L/C, and R/C) at the following three different gap
sizes: (1) 2 ms above each individualÕs gap-detection
threshold (as determined in Experiment 2), (2) 20 ms,
and (3) 50 ms. Thus there were 9 (3 · 3) condition/
gap-size combinations. These combinations were presented in a random order for each listener. The lead/
lag time was fixed at 2 ms, which was well below the
echo threshold for each of the listeners.
After 5 stimulus presentations in each of the 9 condition/gap-size combinations, the listeners were asked to
report their impressions about the gap that occurred in
the middle of the noise by selecting an answer from
the following 6 options: (1) a single gap, (2) a sudden
burst of noise, (3) both a single gap and a noise burst,
(4) two gaps, (5) two noise bursts, or (6) no change.
They were then asked to report which loudspeaker(s)
delivered the perceived gap(s) and/or which loud-
All the 11 listeners reported that they perceived one
or two sudden changes in the middle of the sound in
all combined conditions. No participant used the ‘‘no
change’’ response. However, one male participant appeared not to follow the instructions appropriately. 4
Thus this participantÕs data were not used. The results
from the other 10 listeners appear in Fig. 5.
As shown in Fig. 5, in Condition L/U, all the listeners
predominately perceived the gap as coming from the
right (lagging) loudspeaker. However, there were 1, 4
and 3 listeners who reported that they perceived an additional gap image as coming from the left (leading) loudspeaker in the near-threshold, 20-ms, and 50-ms
conditions, respectively. There were also 2 listeners
reporting that they perceived a noise burst image as
coming from the left loudspeaker in the 50-ms condition. Hence, even though the left and right noises were
never fused, occasionally perceptual events that were initiated by a gap in the right (lagging) sound were attributed to the leading sound. However, for all gap
3
Instructions to Listeners for Experiment 3: ‘‘After you press the
middle button, you will hear noise presented over the loudspeakers.
Listen to the noise carefully because after 5 presentations of the noise,
you will be asked to answer the following three questions: Question 1:
Did you perceive in the noise, (1) a gap of silence, (2) a sudden burst of
noise, (3) both a gap and a noise burst, (4) two gaps, (5) two noise
bursts, (6) no change. Question 2: For the perceived gap(s) in the noise,
please report where the gap(s) came from: (1) the left-hand loudspeaker, (2) the right-hand loudspeaker, (3) the two loudspeakers.
Question 3: For the perceived noise burst(s), please report where the
noise burst(s) came from: (1) the left-hand loudspeaker, (2) the righthand loudspeaker, (3) the two loudspeakers’’.
4
For all condition/gap-size combinations, this listener selected the
same option number, which indicated that his unvarying response was
‘‘noise bursts’’/‘‘two loudspeakers’’.
242
L. Li et al. / Hearing Research 202 (2005) 235–247
Fig. 5. Summary of listenersÕ perceptions of the gap in Conditions L/U (left leading/uncorrelated), L/C (left leading/correlated), and R/C (right
leading/correlated). The gap was only in the sound from the right loudspeaker. The ordinates represent the numbers of listeners, who attributed a
‘‘gap’’ or a ‘‘noise burst’’ to a particular (left or right) loudspeaker at each of the three different gap sizes. Lighter bars indicate ‘‘Gap’’ responses, and
darker bars indicate ‘‘Noise burst’’ responses.
durations, a gap was always perceived in the lagging
sound.
In Condition L/C, the listeners predominately perceived a change in the sound coming from the left (leading) loudspeaker, even though the gap appeared only in
the right (lagging) loudspeaker. When the gap size was
near threshold, all the listeners reported that they perceived only a single gap image in the sound from the left
loudspeaker. When the gap size was 20 or 50 ms, most
listeners perceived either a gap or a noise-burst image
as coming from the left loudspeaker. Only a small number of listeners reported that they perceived a gap or a
burst image as coming from the right loudspeaker.
Hence, when the gap is in the lagging sound and the
sounds are correlated, listeners tend to incorporate any
perceptual change occasioned by the gap into the fused
image, which is perceived to be located on the leading
side. In other words, perceptual changes evoked by a
gap in the lagging sound are captured by the leading
sound. It is interesting to note, that at the longer gap
durations, listeners sometimes heard a noise burst,
which they attributed (with one exception) to the leading
stimulus. One possible explanation for this perception is
that if the gap in the lagging stimulus is long enough,
there is no location information coming from the lagging stimulus to suppress, and the circuitry responsible
for the suppression of location information is disengaged. Consequently, when the gap is terminated, the
L. Li et al. / Hearing Research 202 (2005) 235–247
perceptual system briefly treats the return of the lagging
correlated stimulus as a new stimulus until it re-establishes the correlation between the leading and lagging
stimulus and suppresses the perception of the lagging
source. It is interesting to note, however, that this noise
burst, rather than being attributed to the lagging stimulus is perceived as originating from the direction of the
leading stimulus. In other words, it appears to be captured by the leading stimulus.
In Condition R/C, all the listeners perceived the gap
as belonging to the right (leading) loudspeaker in the
near-threshold condition. At the larger gap durations
(20 and 50 ms), the listeners predominately perceived
the gap (when it was heard as a gap) as belonging to
the right (leading) loudspeaker, but they also reported
hearing a noise-burst image as coming from the location
of the left (lagging) loudspeaker. When there is a gap in
the leading stimulus, there is no leading sound present to
suppress the information as to the location of the lagging stimulus. Hence, one might expect to hear a brief
noise burst during the gap from the location of the lagging stimulus until the perception of the lagging stimulus
is suppressed. This is what appears to have happened
here.
5. Experiment 4
To examine how the precedence effect modulates cortical responses to the probe gap, in Experiment 4, N1,
P2, and long-latency sustained components of ERP
responses to gaps were measured in Conditions L/U,
L/C, and R/C, respectively.
5.1. Method
5.1.1. Participants
All the 11 listeners from Experiment 3 and 1 new
male young university student (21 years old) with normal and balanced pure-tone hearing participated in this
physiological experiment. These listeners were instructed
to remain awake and keep their eyes open, while they listened to the acoustic stimuli.
5.1.2. Apparatus and materials
The apparatus and materials were same as in previous
experiments. However, this ERP recording experiment
was conducted in a different IAC sound-attenuated
chamber that was equipped with 64-channel NeuroScan
SynAmps (bandpass 0.05–50 Hz; 250 Hz sampling rate).
5.1.3. Procedure
The size of the gap in the sound from the right loudspeaker was fixed at 50 ms and the delay between the
sounds from the two loudspeakers was fixed at 2 ms.
243
During the recording, all electrodes were referenced to
the Cz site; for data analysis, they were re-referenced
to an average reference. The analysis epoch included
200 ms of pre-stimulus activity and 3500 ms of poststimulus activity following each of the 150 sound presentations for each of the three conditions: Conditions L/U,
L/C, and R/C. Trials contaminated by excessive peakto-peak deflection (±150 lV) at the electrodes not
adjacent to the eyes were automatically rejected. ERP
waveforms were then averaged separately for each site
and conditions, and digitally low-pass filtered to attenuate the components with frequencies above 12 Hz.
Although the number of stimulus-presentation trials
was 150, the number of trials included in the average
for each condition varied between listeners with the
across-listener average being 116, 114, and 113 for Condition L/U, Condition L/C, and Condition R/C, respectively. For each individual average, ocular artifacts (e.g.,
blinks and lateral movements) were corrected by means
of ocular source components using the Brain Electrical
Source Analysis (BESA) software (Picton et al., 2000).
ERP waveforms were quantified by computing mean
values in selected latency regions, relative to the mean
amplitude of the 200 ms pre-stimulus activity. All amplitude measurements were subjected to mixed ANOVA
with condition and electrode as the two within-subject
factors. Topographic voltage maps were examined using
the 61 electrodes (the periocular electrodes were not
included).
5.2. Results
For the 9 central electrode sites (FC1, FCz, FC2, C1,
Cz, C2, CP1, CPz, and CP2), there were no differences
across these three conditions both for N1-P2 peakto-peak amplitudes to sound onset (F2,22 = 0.238,
MSE = 7.948, p = 0.790) and for slow sustained potentials
following
sound
onset
(F2,22 = 1.308,
MSE = 1.537, p = 0.290) (Fig. 6). However, the N1-P2
responses to the gap did differ significantly across these
three conditions (F2,22 = 9.129, MSE = 3.586, p = 0.001)
(Fig. 6). Pairwise comparisons indicate that the amplitude of N1-P2 response to the gap in Condition L/U
was significantly smaller than that in Condition L/C
(p = 0.022) and that in Condition R/C (p = 0.000), but
the difference between Condition L/C and Condition
R/C was not significant (p = 0.125). Topographic voltage maps for the N1 component to the gap (Fig. 7) indicate that in Condition L/U, the highest negativity was
widely distributed over the midline, but in both Condition L/C and Condition R/C, it became more concentrated over the right hemisphere. Hence, when a gap is
introduced, the cortical response depends upon whether
or not the two sounds were correlated or uncorrelated.
Moreover, there appeared to be ERP differences
in the sustained responses following the gap (Fig. 6)
244
FC1
L. Li et al. / Hearing Research 202 (2005) 235–247
P2
FCz
FC2
C1
Cz
C2
CP1
CPz
CP2
N1
P2
SP
N1
SP
2.5 µV
L/U
-200
1500
L/C
R/C
3500
Time (ms)
Fig. 6. The whole course of the averaged ERP responses recorded
from the central 9 electrode sites across 12 listeners, in each of the L/U,
L/C and R/C conditions. The N1 peak and P2 peak responses to the
sound onset and gap, and slow sustained potentials (SPs) following
sound onset and the gap are indicated in the panel for FC1 electrode
site. The two arrows above the time base indicate the onset of the
sound and onset of the gap, respectively.
L/U
L/C
R/C
Fig. 7. The ERP topographic voltage map for the N1 response to the
gap across the 61 scalp electrodes in each of the L/U, L/C and R/C
conditions.
between conditions. The average amplitude of the sustained responses 550–850 ms after the gap onset was
analyzed. The results show that there were no significant
differences in sustained responses for the two conditions
(L/U and R/C) where the gap was correctly assigned to
the right loudspeaker. However, the condition, in which
the gap in the right sound was perceptually captured
by the left sound (Condition L/C), differed significantly
both from Condition L/U across all the 9 central sites
(p = 0.001) and from Condition R/C across the 3 frontocentral sites (FC1, FCz, FC2) (p = 0.016). Hence, a longlatency and negatively shifted sustained response in the
frontal cortical region following the gap appears to be
associated with gap capture.
6. Discussion
Most previous studies of the precedence effect have
used clicks or short noise bursts as acoustic stimuli to
avoid or reduce the overlap between the leading and lagging stimuli. Here long-lasting sound segments were
chosen for 3 major reasons: First, long-duration sound
segments (e.g., speech or music) are more prevalent in
everyday environments, therefore have greater ecological validity than idealized brief sounds for humans. Second, the use of longer stimuli allowed us to easily
present an attribute (a gap) that was clearly a feature
that appeared only in the lagging sound. Third, only
when the sound duration is sufficiently long, can neurophysiological responses, such as ERPs, to a transient
probe attribute embedded in the sound, be easily distinguished from those to sound onsets and offsets, and the
development of sustained neurophysiological responses
between transient acoustic events be segregated.
In the present study, when the two spatially separated
long-lasting noise sounds were correlated, only a single
noise image was perceived as coming from the location
of the leading loudspeaker if the lead/lag time was below
echo threshold. These results are in agreement with
previous reports that two correlated long-lasting
speech-spectrum noise sounds, which are presented by
two spatially separated loudspeakers (60 separation, 4
ms delay time), can be perceived as a single noise image
originating from the position of the leading loudspeaker
(Freyman et al., 1999).
The average echo threshold found here (9.5 ms), is
within the range reported in previous studies (5–10 ms,
for a review see Litovsky et al., 1999). Interestingly,
when both leading and lagging sounds have comparable
gaps, the gap capture threshold is 15.6 ms, which is significantly longer than the echo threshold obtained from
the same listeners. Hence, for delays larger than 10 ms,
listeners perceive two sound images: one from the leading and one from the lagging loudspeaker. But, as long
as the delays were less than 15 ms, listeners perceived the
gap in the lagging stimulus as occurring in the leading
stimulus, and heard the lagging stimulus as a continuous
noise (no gap). Hence, in the delay region between 10
and 15 ms, listeners hear two spatially separated noises
with the gap belonging to the leading stimulus. These
different capture thresholds (echo versus gap) imply that
different processes are involved in capture for different
attributes.
When the two long-duration noise sounds are uncorrelated, neither the lagging noise sound nor the gap in
the lagging sound is captured. Our results thus lay
emphasis on the importance of inter-sound correlation
in producing perceptual fusion for long-duration
sounds. This notion is partially in agreement with a previous study by Perrott et al. (1987), who used 50-ms
broadband free-field noise bursts (0.2 ms rise/fall, left/
right 20 separation) as stimuli and investigated listenersÕ experience of correlated or uncorrelated noise bursts
at various inter-stimulus onset delays. Perrott et al. reported that fusion was stronger when the two short
noise bursts were correlated than when the two bursts
were uncorrelated. However, when the two noise bursts
L. Li et al. / Hearing Research 202 (2005) 235–247
were uncorrelated and the delayed was below 8 ms, there
were also a small proportion of trials on which fusion of
the two bursts was perceived. In our experiments the
two uncorrelated sounds did not fuse. On a few occasions, however, a gap that appeared only in the right
(lagging) uncorrelated sound was also attributed to the
leading sound, indicating that attributes of the lagging
sound may occasionally be captured by the leading
sound even when the two sounds are uncorrelated.
Hence, although listeners never reported that the two
independent sounds became fused, there is some indication of attribute capture by the leading sound. The disagreement concerning fusion between our data and
those reported by Perrott et al. (1987) for uncorrelated
noises may be due to the differences of stimulus parameters between the two studies, such as those in sound
duration (50 ms vs. 3050 ms), onset/offset duration
(0.2 ms vs. 30 ms), and loudspeaker separation (±20
vs. ±45), etc.
In a reverberant environment, each sound reflection
comes from a location that is usually different from that
of the sound source, and not all attributes of reflections
are suppressed by their sound sources (Clifton et al.,
2002; Freyman et al., 1998; Perrott et al., 1987; Tollin
and Henning, 1999). In the present study, if the gap
attribute in the lagging sound had been suppressed by
the correlated leading sound when the precedence effect
occurred, the gap detection threshold in Condition L/C
should have been higher than those in Condition L/U
and Condition R/C, and the gap detection threshold in
Condition R/C should have been lower than that in
Condition L/U. However, our data show that gap detection thresholds were independent of whether the gap
was in either the leading or lagging sound, and also independent of whether or not the leading and lagging
sounds were correlated. These results are consistent with
the hypothesis that gap detection depends primarily on
the detection of an energy change in the ear on the side
of the loudspeaker producing the gap. On the other
hand, when the two sounds were correlated, a single
compact sound image was perceived as coming from
the leading side; when the two sounds were not correlated, more diffused sound images were perceived as
coming from the both sides. Since there was no difference in gap detection between Conditions L/U, L/C,
and R/C, there is no evidence in this experiment that
sound-image compactness/diffuseness affects gap
detection.
If information in these reflections is not being suppressed, then it has to be somehow perceptually incorporated into the fused image. The present study shows that
when the two sounds are uncorrelated, the lagging
sound is by and large not treated as the reflection of
the leading sound by the auditory system, and two distinct noise images, coming from different directions are
perceived, and the gap presented in the lagging sound
245
is ‘‘correctly’’ perceived as coming from the lagging
loudspeaker. The only exception to this statement is that
sometimes, especially at the longer gap durations, the
gap is also attributed to (captured) by the leading stimulus. In contrast, when the two sounds are correlated,
the lagging sound is treated as a reflection of the leading
sound, a single noise image is perceived, and attributes
that appear only in the lagging sound are attributed to
(captured by) the leading sound. This is not what we
would expect on the basis of the physical cues to the
location of the gap that are present when there is a
gap only in the lagging sound. Fig. 4 shows that when
there is a gap in the lagging (right side) source only,
there is a corresponding drop in energy (especially in
the high-frequency region) in the right ear, with little
evidence of any change in the left ear. Hence, if the location of the gap were to be based on the ear with the most
salient cues, one would expect the gap to be heard on the
side of the lagging sound. Nevertheless, the gap is heard
as occurring on the leading side. In other words, it is
attributed to (captured by) the leading stimulus.
When the gap is only in the correlated lagging sound,
the acoustic situation is ecologically anomalous, because
the gap in the reflection should have its origin in the
source. The ecological prediction is that a gap in the lagging sound would cause a temporary breakdown in the
precedence effect. When the lagging loudspeaker becomes silent during the gap, there is no correlated signal
coming from the lagging loudspeaker to be captured.
Thus, when the gap terminates and the lagging loudspeaker is turned on again, the participant should initially perceive a new sound originating from the
location of the lagging loudspeaker until the precedence
of the leading sound is re-established. However, most of
our listeners did not hear any sound change as coming
from the location of the lagging loudspeaker. Rather
they heard a gap or a burst-like image as coming from
the leading loudspeaker. Since there is no physical gap
in the sound from the leading loudspeaker, the gap in
the sound from the lagging loudspeaker has no leading
‘‘partner’’ to ‘‘fuse’’ with. Moreover, hearing a gap or
a burst-like image as coming from the leading loudspeaker cannot be caused by a peripheral effect, since
there are no obvious differences in the sound spectra
at the left ear (the ear on the side of the leading loudspeaker) between the condition when there is no gap
in the lagging (right-side) stimulus versus when there is
a gap in the lagging stimulus (see Fig. 4). Thus the shift
of gap image from the lagging loudspeaker to the leading loudspeaker denotes the maintenance of the precedence effect during the period of the gap, and must
involve a higher-order attribute capturing process.
On the other hand, when the gap is only in the leading sound that is correlated with the lagging sound, the
acoustic situation is also ecologically anomalous, because a gap in a natural sound source will also appear
246
L. Li et al. / Hearing Research 202 (2005) 235–247
in its reflections. Hearing a gap as coming from the leading loudspeaker and simultaneously a burst-like image
as coming from the lagging loudspeaker indicates a transient disappearance of the precedence effect during the
gap.
Our electrophysiological results suggest a tight link
between subjective perception of the gap and neural responses to the gap. Surprisingly, there is no difference in
ERP responses between the correlated and uncorrelated
sound conditions until a gap occurs, even though the
perceptual responses to the correlated and uncorrelated
noise sounds are quite different. When the two longduration sounds are correlated, the N1-P2 peak-to-peak
response to the gap is enhanced and the N1-topographic-voltage map for the gap shifts laterally towards
the right hemisphere, regardless of the gap being in the
lagging or leading sound. Also, in the frontocentral region, a negatively shifted sustained ERP response following the gap embedded only in the lagging sound
appears to be associated with the perceived capture of
the gap. The present neurophysiological results suggest
that there is a greater need for cortical involvement to
maintain fusion of leading and lagging sounds when
there is a break in one or the other, than to establish fusion at sound onset. This long-latency neural event following the occurrence of the gap also suggests that
higher-order central processes are involved in attribute
capture.
Clinical studies in humans suggest that both the cortex and the inferior colliculus are essential for the precedence effect. Cornelisse and Kelly (1987) reported that
patients with lesions of the right temporo-parietal cortex
were able to localize single clicks but could not localize
the ‘‘fused’’ image of two spatially separated clicks,
when the leading click was delivered from the left hemifield and the lagging click was delivered from the right
hemifield. Litovsky et al. (2002) reported that a patient
with lesions of the right inferior colliculus had substantially weaker echo suppression when the leading sound
was delivered in the left hemifield. Hence it would be
interesting to investigate attribute capture in patients
with unilateral lesions of the central auditory system.
In summary, based on the data of the present study,
three important features of attribute capture should be
noted:
(1) Top-down higher-order processes are involved in
attribute capture. A probe gap introduced in the
leading stimulus can temporarily break the precedence effect whereas introducing a comparable
gap in the lagging stimulus does not break the precedence effect in the majority of our listeners, even
though both situations are ecologically anomalous.
In addition, gap capture is associated with longlatency negatively-shifted slow potentials in the
frontal area.
(2) Attribute capture is not an all-or-none process. For
lead/lag delays between 9 and 15 ms, the location
information concerning the lagging sound is not
suppressed by the leading sound (a sound is still
heard as coming from the direction of the lagging
sound), but a gap in the lagging sound is, nevertheless, captured by the leading sound (a gap is heard
in the leading sound but not in the lagging sound).
This indicates that capture thresholds can differ for
different attributes of the reflection (e.g., gaps in the
lagging sound are more easily captured than other
aspects of the sound). One may speculate that the
degree to which the listener assigns spatially separate and distinct images to the leading and lagging
sounds will depend on the extent to which different
attributes of the lagging sound are incorporated
into (captured) by the leading sound. According
to this speculation, all of the attributes of the reflection would have to be captured in order for the listener to perceive only a single source.
(3) The introduction of a distinct feature such as a gap
into a direct or reflected wave may be one way of
probing cortical involvement in the precedence
effect. In our study, identical ERP responses were
elicited by both correlated and uncorrelated noises,
even though listeners perceive correlated noises to
be quite distinct from uncorrelated noises. One
may speculate that the differences between the two
are processed primarily by brain-stem mechanisms.
However, the ERP to a gap differed substantially
depending upon whether or not the noises were correlated. This suggests that while cortical involvement may not be necessary to distinguish between
correlated and uncorrelated noises, it may be
required to maintain and/or re-establish the perception of these two kinds of noise (especially, with
respect to percepts related to precedence) once there
is a break in either the leading or lagging noise. The
use of gaps as probes may be a way of accessing the
cortical mechanisms involved in the maintenance of
percepts when there are sudden or unexpected
changes in the sensory input. Thus, in order to
more completely understand the neural mechanisms involved in the precedence effect, cortical
neural correlates should be investigated in addition
to the brainstem mechanisms.
Acknowledgements
We thank Jane W. Carey and Neda Chelehmalzadeh
for their assistance during data acquisition. We also
thank the following people who have reviewed previous
versions of the manuscript and provided helpful comments and suggestions to improve its quality: Ann Clock
L. Li et al. / Hearing Research 202 (2005) 235–247
Eddins, William M. Hartmann, Jack B. Kelly and three
anonymous reviewers. This work was supported by the
Natural Sciences and Engineering Research Council of
Canada, the Canadian Institutes of Health Research,
the Canada Foundation for Innovation, and the Ontario
Innovation Trust Fund.
References
Blauert, J., 1997. Spatial Hearing. MIT Press, Cambridge, MA.
Clifton, R.K., 1987. Breakdown of echo suppression in the precedence
effect. J. Acoust. Soc. Am. 82, 1834–1835.
Clifton, R.K., Freyman, R.L., 1989. Effect of click rate and delay on
breakdown of the precedence effect. Percep. Psychoph. 46, 139–
145.
Clifton, R.K., Freyman, R.L., Meo, J., 2002. What the precedence
effect tells us about room acoustics. Percep. Psychoph. 64, 180–
188.
Clifton, R.K., Freyman, R.L., Litovsky, R.Y., McCall, D., 1994.
ListenersÕ expectations about echoes can raise or lower echo
threshold. J. Acoust. Soc. Am. 95, 1525–1533.
Cornelisse, L.E., Kelly, J.B., 1987. Neuropsychologia 25, 449–452.
Fitzpatrick, D.C., Kuwada, S., Batra, R., Trahiotis, C., 1995. Neural
responses to simple, simulated echoes in the auditory brainstem of
the unanesthetized rabbit. J. Neurophysiol. 74, 2469–2486.
Fitzpatrick, D.C., Kuwada, S., Kim, D.O., Parham, R., Batra, R.,
1999. Responses of neurons to click-pairs as simulated echoes:
auditory nerve to auditory cortex. J. Acoust. Soc. Am. 106, 3460–
3472.
Freyman, R.L., Clifton, R.K., Litovsky, R.Y., 1991. Dynamic
processes in the precedence effect. J. Acoust. Soc. Am. 90, 874–884.
Freyman, R.L., McCall, D.M., Clifton, R.K., 1998. Intensity discrimination for precedence effect stimuli. J. Acoust. Soc. Am. 103,
2031–2041.
Freyman, R.L., Helfer, K.S., McCall, D.D., Clifton, R.K., 1999. The
role of perceived spatial separation in the unmasking of speech. J.
Acoust. Soc. Am. 106, 3578–3588.
Hartung, K., Trahiotis, C., 2001. Peripheral auditory processing and
investigations of the precedence effect which utilize successive
transient stimuli. J. Acoust. Soc. Am. 110, 1505–1513.
Levitt, H., 1971. Transformed up-down methods in psychoacoustics. J.
Acoust. Soc. Am. 49, 467–477.
Li, L., Yue, Q., 2002. Auditory gating processes and binaural
inhibition in the inferior colliculus. Hear. Res. 168, 113–124.
Liebenthal, E., Pratt, H., 1999. Human auditory cortex electrophysiological correlates of the precedence effect: binaural echo lateralization suppression. J. Acoust. Soc. Am. 106, 291–303.
247
Litovsky, R.Y., 1998. Physiological studies on the precedence effect in
the inferior colliculus of the kitten. J. Acoust. Soc. Am. 103, 3139–
3152.
Litovsky, R.Y., Delgutte, B., 2002. Neural correlated of the precedence
effect in the inferior colliculus: effect of localization cues. J.
Neurophysiol. 87, 976–994.
Litovsky, R.Y., Shinn-Cunningham, B.G., 2001. Investigation of the
relationship among three common measures of precedence: fusion,
localization, and discrimination suppression. J. Acoust. Soc. Am.
109, 346–357.
Litovsky, R.Y., Yin, T.C.T., 1998a. Physiological studies of the
precedence effect in the inferior colliculus of the cat: I. Correlates of
psychophysics. J. Neurophysiol. 80, 1285–1301.
Litovsky, R.Y., Yin, T.C.T., 1998b. Physiological studies of the
precedence effect in the inferior colliculus of the cat: II. Neural
mechanisms. J. Neurophysiol. 80, 1302–1316.
Litovsky, R.Y., Rakerd, B., Yin, T.C.T., Hartmann, W.M., 1997.
Psychophysical and physiological evidence for a precedence effect
in the median sagittal plane. J. Neurophysiol. 77, 2223–2226.
Litovsky, R.Y., Colburn, H.S., Yost, W.A., Guzman, S.J., 1999. The
precedence effect. J. Acoust. Soc. Am. 106, 1633–1654.
Litovsky, R.Y., Fligor, B.J., Traino, M.J., 2002. Functional role of the
human inferior colliculus in binaural hearing. Hear. Res. 165, 177–
188.
Perrott, D.R., Strybel, T.Z., Manligas, C.L., 1987. Conditions under
which the Haas precedence effect may or may not occur. J. Audit.
Res. 27, 59–72.
Picton, T.W., van Roon, P., Armilio, M.L., Berg, P., Ille, N., Scherg,
M., 2000. The correction of ocular artifacts: a topographic
perspective. Clin. Neurophysiol. 111, 53–65.
Rakerd, B., Hartmann, W.M., Hsu, J., 2000. Echo suppression in the
horizontal and median sagittal planes. J. Acoust. Soc. Am. 107,
1061–1064.
Shinn-Cunningham, B.G., Zurek, P.M., Durlach, N.I., 1993. Adjustment and discrimination measurements of the precedence effect. J.
Acoust. Soc. Am. 93, 2923–2932.
Tollin, D.J., Henning, G.B., 1999. Some aspects of the lateralization of
echoed sound in man. II. The role of the stimulus spectrum. J.
Acoust. Soc. Am. 105, 838–849.
Wallach, H., Newman, E.B., Rosenzweig, M.R., 1949. The precedence
effect in sound localization. J. Acoust. Soc. Am. 62, 315–336.
Yin, T.C.T., 1994. Physiological correlates of the precedence effect and
summing localization in the inferior colliculus of the cat. J.
Neurosci. 14, 5170–5186.
Zurek, P.M., 1980. The precedence effect and its possible role in the
avoidance of interaural ambiguities. J. Acoust. Soc. Am. 67, 952–
964.
Zurek, P.M., 1987. The precedence effect. In: Yost, W.A., Gourevitch,
G. (Eds.), Directional Hearing. Springer-Verlag, New York, pp.
85–105.